Local AI

When Google's Agent Calls, Does It Pick Your Shop?

Published October 23, 2026

By Aaron Roberts · 6 min read

A customer asks Google to find someone who can fix a leaking water heater in Horseheads. The request includes a deadline, a budget, and one condition: somebody needs to answer.

That's a hypothetical customer. But Google calling businesses to check pricing and availability is a real feature, not a prediction.

The question for your shop is getting bigger than where your website ranks. Can a system find you, understand whether you fit the job, and get a usable answer when it contacts you?

You can have a good website and still lose at that last step.

What Google actually said

In an interview covered by Search Engine Journal, Sundar Pichai described an "agentic layer" connecting Google's products. His point was that an assistant could carry a task and its context across those products, rather than making the customer start over each time.

That sounds abstract. The practical version is simpler: less time finding information, more help getting something done.

Google already announced AI-powered calls to local businesses for pricing and availability in July 2025. At I/O 2026, it described expanded calling for select categories, including home repair, beauty, and pet care, alongside booking features that provide links to finish with a provider.

Google also reported more than 2.5 billion monthly active users for AI Overviews at I/O 2026. That's the audience for its AI summaries. It DOESN'T mean 2.5 billion people are using agents to book contractors.

Keep that distinction. An AI summary, an AI phone call, and a completed booking are three different things.

Getting considered isn't the same as getting the job

Here's an example, not a claim about Google's ranking formula.

Two plumbing companies serve the same town. Both have websites. Both show up when you search their names.

The first says it handles emergency water-heater repairs, lists its service area, and has current hours. When someone calls, the person answering can explain the next step and whether a technician might be available.

The second says "quality solutions for all your needs." Its hours disagree with its business profile. Calls go to a full voicemail box.

If you're trying to arrange a repair, one company gives you something to work with. The other gives you more work.

That's true whether the caller is a homeowner, an office assistant, or software working for the customer.

We don't have a published formula that tells us how Google's agents weigh every local business. Anyone selling a guaranteed spot as "the one the agent picks" owes you evidence.

But you don't need that formula to fix a dead phone number or explain what you actually do.

Make the important facts easy to confirm

Start with the questions your front desk already gets:

Do you handle this specific repair?
Do you serve this town?
Are you taking new customers?
What happens after I request service?
Is there a diagnostic or call-out fee?
Does after-hours mean a live answer, a callback, or a technician on the road?

Put the stable answers on your website in normal sentences. Keep your business profile consistent with them.

Don't publish a promise your crew can't keep. "Emergency service" needs an explanation. If you answer after hours but schedule the visit the next morning, say that.

Availability changes. A website doesn't need to pretend it's a live dispatch board. It needs to make the next step clear and route the inquiry to somebody who can confirm it.

Then test what happens when the phone rings

Your website can get the business considered. Your intake process still has to finish its part.

Call your own number from another phone. Try it during lunch. Try it after the office closes. Follow the same steps a new customer would follow.

Can the person answering identify the service area? Can they capture the job details? Do they know what they're allowed to promise? If they need help, where does the caller go?

A voicemail greeting that says "we'll call you back" might be accurate. But it doesn't answer a time-sensitive availability question. Decide how your business handles that situation before it happens.

If you use an AI receptionist, the same limits apply. It needs current information, clear escalation rules, and a way to avoid inventing a price or an appointment. Booking only works when the required calendar or dispatch integration is configured and working.

Two agents talking to each other won't fix a business process nobody wrote down.

Count more than website visits

A call made on a customer's behalf might not come with a website visit you can trace. So website traffic alone won't tell you whether this channel is producing useful inquiries.

Add a few fields to your intake notes:

How did the customer find you, if known?
Was an automated assistant involved?
What service and location did they ask about?
Did they get a clear next step?
Did the inquiry turn into a booked job?

Record what you know. An unknown source stays unknown. Don't label every unfamiliar caller a Google agent.

Over time, those notes can show whether the problem is getting found, getting contacted, or getting the customer scheduled. Those are different fixes.

What I'd do this week

Check your business profile against your website. Fix conflicting hours, service areas, and phone numbers.

Read your main service page as if you've never heard of the company. Can you tell what it repairs and where it works?

Then make the test call. Write down what happened instead of assuming the office handled it.

None of that guarantees an AI recommendation. It does remove problems that can cost you an ordinary customer today.

I think that's the useful way to prepare for agent-driven search. Make your business easy to understand and your intake process reliable enough to act on.

Google may help the customer make the call. Your shop still has to give them a reason to take the next step.

If you're not sure where that process breaks, our AI Assessment is a place to start. We can review how inquiries reach your business and where automation would help, without promising a ranking nobody controls.

Sources

Aaron Roberts runs Fuzzy Solutions, LLC in Elmira, NY. He builds AI phone agents and lead-gen websites for service businesses across the Southern Tier.

← All posts